Starlight Mini stops using GPU mid-encode (GPU usage drops to 0%) – reproducible issue

Topaz Video Bugs and Issues
1

Hello Topaz Labs Support Team,

I would like to report a reproducible issue with Topaz Video AI when using Starlight Mini, where GPU processing stops during encoding even though the job appears to continue.

Environment

  • Product: Topaz Video AI

  • Version: 1.2.0

  • AI Model: Starlight Mini

  • GPU: NVIDIA RTX 5070 Ti

  • VRAM: 16 GB

  • OS: Windows 11

  • Driver: Latest NVIDIA Studio/Game Ready driver

  • Input video: FHD (1920×1080), 30 FPS

Issue Description

When encoding with Starlight Mini, the process starts normally and GPU usage is initially observed. However, during the same encoding job:

  • GPU usage suddenly drops to 0% (confirmed via NVIDIA App and Windows Task Manager).

  • Topaz Video AI continues to display “Processing” and does not report an error.

  • Encoding speed effectively stalls or becomes extremely slow.

  • GPU usage does not recover for that job.

Restarting Windows temporarily resolves the issue, but the problem frequently reoccurs during the next encode, especially with longer videos or multiple encodes in one session.

Observations

  • The issue occurs even with FHD input and sufficient VRAM.

  • VRAM is not fully exhausted when the problem happens.

  • Restarting Topaz Video AI alone does not fix the issue once it occurs.

  • Restarting Windows resets the behavior temporarily.

  • Switching to other models (e.g., Artemis / Proteus) does not trigger the problem as often.

  • This appears consistent with a CUDA context or memory handling failure specific to Starlight Mini.

Expected Behavior

GPU utilization should remain active throughout the entire encoding process, or the application should fail gracefully with an error if GPU processing cannot continue.

Suspected Cause

Based on behavior, this may be related to:

  • CUDA context loss or improper cleanup

  • GPU memory allocation / deallocation instability

  • A driver-level GPU watchdog interaction triggered by Starlight Mini

Temporary Workarounds

  • Restarting Windows between encodes

  • Processing videos in short segments (5–10 minutes)

  • Avoiding batch processing

  • Reducing Starlight Mini feature load

Request

Could you please confirm:

  1. Whether this is a known issue with Starlight Mini in version 1.2.0

  2. If there is a recommended stable configuration or upcoming fix

  3. Whether debug logging would be helpful for further diagnosis

Thank you for your time and for your continued work on Topaz Video AI.

2

We will need to review the logs from the app to see what is going on here, along with your system profile report. Can you pull those and send to the support team at help@topazlabs.com

How To Obtain Your System Profile (Windows)

To gather logs, please select Help > Logging > Get Logs for Support and attach the zip file to your reply.

Here is a video to help with the steps of how to collect the logs.